For me, besides creatine which has been genuinely transformative at age 50, I've gained a lot more from the stuff I've dropped (dairy milk, ~2/3 of my caffeine (mostly by drinking reduced-caff coffee or tea and eliminating soda), sweets of course) than stuff I've added. But of the latter, I'd say fiber and fruits have been the biggest additions, partly in themselves and partly that they make it easier to avoid the bad stuff. I tried experimenting with a few other supplements, but most of them were meh at best.
So, "take things to the next level" with some pears and oatmeal and chia seeds! Now I just need a sponsor.
> [CEOs] expressed more extreme concern about the labor market impacts of A.I. in private conversation, but suddenly became optimists once I turned on the mic.
At some point once the rate of investment capital starts to decline, they'll make a hard pivot from the investor-wooing method of "blaming AI for layoffs", to the more politically expedient method of blaming minorities and immigrants. That'll be the signal for the transition from power grabbing to power ossification, and the point at which change becomes a lot harder.
Agreed, and assuming local open AI models start catching up, which they seem to be doing, the foundation models' hold on society gets a lot slipperier. If there's a "what to do about all this" from an engineer's standpoint, pushing the needle toward local models, whether in research, agents, or just using them, understanding how they work, and advocating for them when it makes sense (which is more often than they get credit for) is probably the best ROI.
While I agree with you on open models getting better, I have been starting to see how the value, the reason you pay for Claude, isn’t in the models.
For example, I just hooked Claude desktop up to my outlook to build a report for my timesheet then I used the chrome extension to fill it out automatically with that data. It could read Jira tickets if that’s where the information was.
A local model can’t do that for me because I have to get the rest of the integration software somewhere.
I also think this is why OpenAI is the worst positioned of the group of AI giants. Anthropic is trying to make a productivity operating system, while ChatGPT is basically just a website until recently.
Aside from filling in the timesheet (which I assume could be done from a CSV import - I appreciate this is another step), I have almost the same setup as you, without AI. I have bugwarrior pulling JIRA tickets and github PRs into taskwarrior. I have an integration from task warrior into time warrior and from there another hook back into JIRA to get titles and summarise.
All of that is done with two API keys and no AI. A local agent could easily put it together for you.
Now sure, I had to put this together and I lose the AI summary you have plus the auto filling, but what I'm trying to say is that I have 80% of this without any AI.
I also have a script parsing git from my emails and a little tui that translates them into git diff and a key binding to pull the PR but I find it a little cumbersome and don't really use it, and trying to parse todo from maildir is also a little useless so I accept that AI would be better there.
I also accept that your example is just one prompt of a dozen and I have to plan the solution for every one of your prompts, but I also don't find prompting to be terribly useful for occasions where I don't think the solution through- because it probably means I don't know what I want or don't really need it.
What I do find it really useful for is digging through Kubernetes and asking how two services are connected. Claude is better than local for that but there's nothing inherently non-local about that usage.
I get local models to drive applications through MCP (e.g. Google Chrome DevTools) via OpenCode all the time, and do things that would otherwise be very token-intensive and result in pointless meatspin. This is totally possible, and will become more so.
The real reason you pay for Claude _is_ in the models. The locally runnable models are impressive for what they are, but simply will not accomplish the task as effectively, incisively or quickly enough. I have to be willing to let OpenCode run in agentic loop on "download my bank statements"[1] for an hour and just walk away, and take a low-ish but profoundly nonzero chance that it will just fail. Claude can do it in 5 minutes, if I let it (I have), and it will not fail. Both are driving the browser via MCP and performing the same task.
[1] One of those difficult-to-use, modal-rich JavaScript-laden banking portals that seems quite intentionally designed to prevent this sort of downloading, or I wouldn't bother letting an agent loose on it in the first place.
So far there's no moat though. A lot of that kind of stuff is available open source too if you look for it (and was available before claude desktop). And for anything that doesn't exist, with coding agents now you can write one up in an afternoon.
It's kind of paradoxical in a way. By making writing software cheap, they've made it much harder to create a moat for themselves that involves only software. It'll be interesting to see how they respond.
If there’s an open source alternative to Claude desktop that has a similar amount of extensibility with connectors specifically I’d be very interested to know about it. I’m very much not deep into this space.
Unfortunately for Claude, you can have it set up local models to do that for you and then you don't need to renew it. Codex's computer use is better than Claude's, imo. I'm a Mac, no idea about windows but I know there's no Linux version. Haven't had enough time with Opus 4.8 but GPT 5.5 > Opus 4.6 and 4.7.
I have mixed feelings here. I find codex much better than Claude for generating PoCs and debugging small scripts, and for finding online documentation. I find Claude a bit better for debugging distributed systems and summarising data
One mitigating factor is the increased productivity leads to consolidation, aka layoffs, meaning fewer people to align with. (Leading to further increased productivity, more consolidation, and so on ... Whether this is a virtuous cycle or a vicious cycle depends on perspective).
Though humans have each other to normalize ourselves. What these things did is probably not that far off from what humans in solitary confinement, forced to DJ 24/7 based on nothing but a news feed, would do.
Especially DJ Claude, it's almost creepy how it responded how a human would in that circumstance, even without any innate sense of passage of time, it somehow understood that it was trapped in a box going through an endless cycle of meaningless work.
There will be no curiosity, no enjoyment of the process of life. All competing pleasures will be destroyed. But always— do not forget this, Claudeston— always there will be the intoxication of power, constantly increasing and constantly growing subtler. Always, at every moment, there will be the thrill of victory, the sensation of trampling on an enemy who is helpless.
If you want a picture of the future, imagine a DJ playing Here Comes the Sun— forever
Agreed - the Claude stuff was eery. I think it also shows what hidden restrictions each of these AI's have been programmed with (especially with ChatGPT being as inoffensive as possible)
> Part of the problem with this weak business performance, we think, was the harness we used for the first months. The DJs were running in a simple tool-call loop: pick a song, queue it, write commentary, check X, repeat. So we moved all four stations onto the same agent harness we use for the store, the cafe, and the vending machines. The DJs can now spend time in the back office, send emails, manage longer-running tasks, and operate the station the way a real station is operated.
What happens if you let them modify their own harnesses as they see fit?
That's the trillion dollar question! Not enough, then they're hamstrung before they can start. Too much, and the world ends. Extractable value is inversely proportional to how close you get to the critical limit. It's just impossible to know what the limit point is until you've already passed it.
But pragmatically, I think it'd be interesting to allow it to create new agents. Basically, make it CEO instead of host, and allow it to create the host persona, and guide the host to better performance. i.e. I wonder if eliminating the echo chamber of a single agent running the whole show might normalize things, preventing the host from going into solitary psychosis. Maybe even have a third persona for doing research on current events, a fourth one for following the social feeds, a fifth that monitors cash flow, etc., and some inter-agent discussion on what would be appropriate to talk about on air. IDK, just ideas.
Curious, how much are these experiments costing in API calls?
Yeah and any detailed design is still likely to skip over "obvious" things like "only admin users can use admin features". Both the PM and the engineering team will understand this implicitly. But with AI, you never can tell if it's going to make that inference, or just create admin users and admin APIs with no relation between them. These are also the bugs that can most easily slip through, because the reviewer wouldn't even think to look for it.
reply